A Wi-Fi FTM-Based Indoor Positioning Method with LOS/NLOS Identification

Round 1

Reviewer 1 Report

The article is very interesting. However, before it is published, it should be corrected as suggested below.

Line 25: please define LBS Line 29-30: some editorial issues with comma Line 36: redefine ILS Line 85-86: pr has two meanings? Is it time or location? What are the units for eq. (1)? I would suggest to use in eq. (1-2) “e” for “error” and “t” for time, since this is more common for readers. Than the indexes could be related to the particular parameters. Eq (4): d and d0 not defined Sections 3.1. The description is not enough for the Figure 2, which is not very informative. Three blocs have different shape but same content. This should be improved. Line 145: rewrite 5HZ into 5 Hz or 5Hz. Same for line 302. Line 146-147: “It can be seen that LOS and NLOS ranging results are similar in any distance, but RSSIs are quite different” this should also be discussed more detailed. Why for similar RSSI los/nlos are very correlated Ranging Results? “The experimenter placed Wi-Fi devices in an open corridor and in a closed room, 144 then collected RSSIs and ranging results at 1 m intervals in the corridor” – which results are for which room? Eq (5) and (6) are exactly the same ??? please delete one of them or improve if there is a typo error. What is RSSI_d in eq (7)? Figure 4. This is a very simple detection method. But still on what dataset the P is calculated? If it is done based on gauss distribution is maybe needs some history dataset? What is than N in eq. (7)? How is “mu” calculated? Line 160-161: “RSSIs were input into the Gaussian model, and the probability could be output.” – there is sth weird in this sentence. Figure 7 c) I would try, on your place, to add these two exponential functions with different amplitudes and scales and fit it again (easy task with Matlab curve fitting or fminsearch). This should give better fitting results. The theoretical curve of RSSI decay with distance is well known. You should include MSE_doubleExpModel(Ranging Result) as Figure 7d). There is something curious in the fact, you define MSE in m and model RSSI in dBm. Should not it be in the same unit?? How do you calculate the MSE? For future I would also suggest to make experiments on longer distances, 24 meters is quite short for WiFi applications. Figure 8: Please add errorbars to the fitting-results. Table 3: Maybe there is a typo error in the Method column? Why do you assume the variance (sigma^2) follows the same exponential model as RSSI does? Maybe other fitting model would be better? The discussion part is quite short. Maybe this is summary or conclusion? And within it: “(…) paper proposed (…) indoor positioning method with LOS/NLOS identification. A LOS/NLOS identification model was proposed. The model was suitable for most indoor environments, (…)” Please rewrite the above mentioned, there are three sentences with the same information. Line 343: doubled “that”

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This work presents Wi-Fi FTM based indoor positioning with LOS/NLOS identification. The method is simple. Also, the experiment has been actually made and the results look fine.

However, all methods used in this paper uses straightforward application of routine procedure that should have been done as such. In other words, any novelty or originality of algorithms and experiments are not found clearly. Therefore, it is strongly encouraged to present novelty and original contributions of the work clearly.    

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I suggest to accept the paper after revision of small typos.

1) Please, mark that there are some editiing problems: "Error! Reference source not found.," in the text!

2) Eq. 9 - both cases are for same range --> change the bottom to d>=15 (?)

Suggestion for future: You may also create your own mathematical functions, as combination (+*/^ etc.) of various basic functions (sin, cos, ln, exp, etc.) and subject them for fitting to your data using Matlab. The in Matlab available functins (e.g. cfittool) do not "fit" to every real life data, thus creating an "own" mathematical model, may give optimal fitting results.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf